Advanced Noise-Resistant Electrogastroenterological Classification Employing Convolutional Neural Networks and Hybrid Wavelet Transform Denoising
Abstract
Analysis of electron gastroenterological signals is one of the urgent issues in the diagnosis and treatment of gastrointestinal diseases. The correct analysis of these signals is significantly affected by various noises and distortions since there is a specific mechanism for obtaining information from living organisms. The main goal of this study is to improve the results of the classification of electron gastroenterological signals by using various noise elimination methods to increase the accuracy of the diagnosis of gastrointestinal diseases. In the study framework, multiple noises were added to electrogastroenterological signals to model real conditions. To eliminate these noises, wavelet transform, median filter, Gaussian filter, and a hybrid method of wavelet and median filters were used. The results of the study showed that the proposed hybrid method is significantly more effective than other methods. The multiscale analysis capabilities of the wavelet transform and the ability of the median filter to preserve important signal features played a key role in achieving these results, and will also prove important in future research and the development of real-time EGG applications.